Autoloom README

The project's README from github.com/GanyuanRan/Autoloom. As one markdown file: README.md.

Autoloom

Governance Engineering for Coding Agents

Free desktop client · Your choice of model · Windows x64 Beta

Latest release, including Beta Windows x64 Documentation in Chinese and English

TypeScript Electron React Aegis Governance

Website · Download Beta · Documentation · Feedback & ideas · 中文

40-second real-task demo

https://github.com/user-attachments/assets/7a5f00f4-2761-452f-a122-0ee6ef01c3f0

Watch on the website →

ConversationSettings
Autoloom Conversation Autoloom Settings

AI coding with engineering judgment built into execution.

Autoloom brings Aegis’s core governance methods into the coding runtime: question whether a change is needed, challenge assumptions before editing, and review delivery against requirements and execution evidence.

Spend less time reminding your agent to investigate first, avoid unnecessary complexity, and show what it actually verified.

Reasoning at the points where decisions matter

Method Its role in the development workflow
First principles and change necessity At task initiation, examine goals, facts, existing capabilities, and constraints to establish why a change is needed and whether an adequate solution already exists.
Anti-entropy and structural complexity checks Before changes, examine duplicated responsibilities, unnecessary layers, and maintenance burden so added complexity addresses a concrete problem.
Counterexample review and causal diagnosis Before changes, examine conditions that could invalidate a solution; after execution failures, require diagnosis grounded in observations and evidence.
Implementation conformance and completion review Before recording results, compare task requirements, actual execution, and verification material, stating what is complete and what remains unverified.

Autoloom delivers these method requirements at the relevant points, checks prerequisite ordering for key operations, and retains process records. Whether a method was applied well and a result is reliable still needs to be judged from code, execution results, and verification evidence.

Before an admitted project mutation runs, the conversation can show a collapsed Governance impact row describing how the relevant methods changed that concrete action. Its user-facing fields follow the request language, and the complete recorded details remain inspectable.

Recorded project goals, constraints, and key decisions stay with the project. Changes to those decisions present a specific diff for your review. You can also configure a test-first preference and code complexity thresholds to reflect your project's requirements.

From engineering methods to execution

Aegis (GitHub) is our open-source AI engineering method pack. Autoloom integrates its core governance methods into development, covering change necessity, complexity, causal diagnosis, and delivery review.

Guidance loads as needed. The runtime maintains prerequisites for key actions, task state, and execution evidence—connecting governance to actual work and reducing repeated reminders and manual checks. Reliable delivery still depends on actual verification results.

Built on DeepSeek Harness, with our own approach to development

Autoloom is built on DeepSeek's official open-source DeepSeek Harness dsh-v0.1.1-rc.2, at source commit b150a551, retaining the Cordis plugin mechanism.

We will continue following official upstream releases and learning from worthwhile improvements. Changes that fit Autoloom and solve real problems will be adopted after adaptation and verification; selected later reliability fixes are already included.

We also maintain our own product direction: bring engineering judgment into automated development, ground the process in evidence, make results inspectable and keep reducing the user's supervision burden. Autoloom has its own version numbers and release cadence.

We thank DeepSeek Harness, Cordis and the related open-source projects for their contributions. The client retains third-party licenses and copyright notices.

Where we are headed

Become a trustworthy, fully automated AI software development platform.

Help ideas progress through requirements, design, implementation, verification, and continued iteration. Make more of that work autonomous while keeping key decisions, supporting evidence, and delivered results inspectable.

That ambition comes with an ongoing question: how can we achieve the best possible balance of cost, speed, and quality?

We care about the total cost of reliably delivering a requirement: model calls, waiting time, human supervision, and rework caused by defects. Reducing unproductive attempts and repeated context, and directing verification effort toward what matters to the result, are continuing priorities. Different tasks need different tradeoffs. We aim to improve how much of all three can be achieved together and test those improvements on real tasks.

Autoloom is in Beta. Fully automated development and a better balance across these three dimensions are goals we are working toward.

macOS and Linux desktop clients are planned, so you can use Autoloom on your preferred operating system. Windows x64 is currently available; progress on macOS and Linux will be announced in this repository.

Follow what comes next

If you want AI development to include this kind of engineering judgment, ⭐ Star Autoloom to support its continued development. Follow the link to the top of the repository, then click Star in the upper-right corner.

For new releases and repository activity, select Watch → All Activity at the top right of this repository. You can adjust notification delivery in GitHub settings. Use Star to bookmark and show support; use Watch to subscribe to notifications.

Get started

  1. Go to Releases and download the Windows x64 installer, Autoloom-<version>-win-x64.exe.
  2. Install and open Autoloom. Add your model provider and API key in settings, then select a default model.
  3. Open a test project you have backed up and start with a clearly scoped change.

The client is free; your chosen provider bills model usage.

The workspace supports local projects and Linux x64 projects connected over SSH, with project files, code diffs, command logs, previews, and docked interactive terminals. Remote environment requirements are listed in the Beta notes below.

Web search works in local and SSH projects through built-in public Parallel/Exa MCP services, without a separate search API key. Direct page reading supports proxy Fake-IP DNS while retaining public-network access checks. Anonymous search remains subject to upstream limits.

The installed client supports background checks and downloads for Beta updates. Once a download is ready, click “Restart to update” to install and reopen the client. Finish or stop active tasks first; ordinary exit does not install an update. Users of the older extracted ZIP must install the first installer release manually.

The user documentation covers model providers, permissions, local and SSH projects, Session backup, updates, and troubleshooting.

Project information

Bring a real task and try it

Tell us through Issues what Autoloom helped you finish, which steps still needed repeated reminders, and where you hesitated to accept its results. Your actual experience will help us choose what is most worth improving next.

When reporting a problem, include the version, system information, reproduction steps, and expected and actual results. Remove keys, access tokens and personal information before sharing logs or screenshots.

LINUX DO

This repository is Autoloom's official hub for downloads, release notes and feedback.

Autoloom’s client source code is not currently public. As the project evolves, we’ll evaluate what to open source and how, taking community feedback and sustainable maintenance into account. The current closed-source approach is not a commitment to staying closed source permanently. We’ll share concrete plans in this repository when they’re ready.

Beta notes and validation scope
  • The current release is a Windows x64 test build. Data formats may change during Beta; back up important projects and task records before upgrading. Incompatible versions will provide manual upgrade instructions.
  • SSH remote projects support Linux x64 and require a working SSH connection and Node.js 22.x, version 22.19 or later. Restricted commands also require a working Bubblewrap sandbox backend.
  • The installer has no Windows code signature, so Windows may show an unknown-publisher or SmartScreen prompt. Download from this repository's Releases and verify the checksum without disabling system security features.
  • Each release provides SHA256SUMS. Check the file with PowerShell using Get-FileHash -Algorithm SHA256 <installer-path>.
  • Consult each Release for its changes and validation scope. Keyless installation, update, or interface checks do not establish that real-model end-to-end tests passed.

Autoloom (GanyuanRan/Autoloom)

beta
Searching Database…

1K ·GitHub
Autoloom
What is this?Autoloom is a desktop AI coding client that bakes Aegis engineering-governance methods right into the agent's execution loop, so code changes are checked for necessity, reviewed for complexity and diagnosed with evidence before being delivered.
Is this a cloud service or something I install?It is a free downloadable desktop app, currently a Windows x64 Beta build, where you bring your own model provider rather than a hosted cloud agent.
How do I get it running?Download the Windows x64 installer from the repo's Releases page, install and open Autoloom, then add your model provider's API key in settings and pick a default model.
Do I need my own API key?Yes, Autoloom itself is free but you must connect your own model provider; the app doesn't charge you, your provider bills for the model calls.
Any install pitfalls or rough edges I should expect?The Windows installer isn't code-signed yet, so you'll likely see a SmartScreen or unknown-publisher warning; verify the release's SHA256 checksum instead of disabling Windows security features, and expect Beta data formats to still change.
What platforms does it run on, and can I work on a remote Linux box?Right now only a Windows x64 client exists, with macOS and Linux clients planned but unreleased; you can still work on remote Linux x64 projects over SSH, which requires Node.js 22.19+ on that remote machine.
What is Aegis?Aegis is the team's separate open-source AI engineering method pack covering change necessity, complexity checks, causal diagnosis and delivery review. Autoloom integrates these methods directly into the coding runtime instead of leaving them as documentation you'd have to remind a coding agent to follow.
What is DeepSeek Harness and why is it mentioned here?Autoloom is built on DeepSeek's open-source DeepSeek Harness runtime (dsh-v0.1.1-rc.2), keeping its Cordis plugin mechanism, while Autoloom's own team layers the governance features on top and tracks useful upstream fixes separately.
Is Autoloom's source code open?No, the client source isn't public right now; this repository only serves as the official hub for downloads, release notes, and feedback, though the team says closed source isn't necessarily permanent.
Can the agent search the web while coding?Yes, built-in Parallel and Exa MCP services provide web search for both local and SSH-connected projects without needing you to supply a separate search API key.
What are similar projects?cocode, zuse, qwen-code, Baize, acryl, claurst, kimi-code, open-swe
Which topics is it in?beta, software-development, agentic-coding, byok, ai-coding, ssh, coding-agents, deepseek-harness, windows-app, coding-agent, ai-agent, agent
Creator
GanyuanRan
Size
12.2 MB
Stars
1K
Updated
CommitOct 17
Age
18d
Commits · 90 Days